Regional-scale correlation between CO2 fire emissions, burned areas, and mid-tropospheric CO2 daily variations over southern Africa
Monthly mean mid-tropospheric CO2 columns are retrieved from evening and morning observations of NOAA-10 satellite over the tropics during the period 1987-1991. We find that the difference between evening and morning CO2 columns (hereafter referred to as Daily Tropospheric Excess - DTE) increases by up to a few ppm over regions affected by fires. A high positive correlation (R2~0.8) is found between annual DTE and CO2 emissions derived from burned area (Global Fire Emission Database - GFEDv2) across 10 regions with contrasted vegetation cover in southern Africa. Seasonal variability comparison between DTE and GFEDv2 also shows a good general agreement. Only two regions south of 10° S, show a seasonal increase of DTE starting earlier and rising up more rapidly than seen in two burned area products: GFEDv2 and L3JRC, the latter established by the Joint Research Center. The phase of the L3JRC dataset is however closer to DTE observations. This misfit could come from limitations in current burned area detection algorithms (difficulty in detecting small fires). 3-D simulations of the DTE signal by the LMDz General Circulation Model, in which a pyro-thermal plume model is activated, confirm the observations. A large fraction of fire products are directly injected in the mid-troposphere, well above the boundary layer. This rapid uplift of CO2, combined with atmospheric transport patterns in southern Africa during the dry season, characterized by a fluctuating continental gyre, produces a daily DTE signal mainly positive above the source region and either positive or negative outside of the source region. On a monthly mean, this results in a persistent DTE signal above the source region of an order of 1 ppm, while the impact of large-scale advection vanishes. We conclude that the DTE signal is a quantitative proxy of fire emission spatial patterns, in particular before the ATSR or MODIS observation periods when better quality fire count and burned area data became available, and can also bring a constraint in the analysis of their present results.